A coevolutionary algorithm using multi-operator ensemble for many-objective optimisation problems

MaOPs are typically solved by using evolutionary algorithms (EAs) to search solutions with the help of operators. The strategy of multiple-operator ensemble (MOE) can combine the search capabilities of different operators to ensure better adaptability in different fitness landscapes. This paper proposes a MaOEA/D algorithm based on coevolutionary multi-operator ensemble (MaOEA/D-CME) for solving MaOPs. The algorithm utilises coevolution technique to balance the capabilities of the simulated binary crossover operator (SBX) and the differential evolution operator (DE) in MOEA/D for different types of problems. To reduce computational costs and avoid premature convergence or slow convergence, we propose a 'multi-stage environmental selection' strategy. Tested on benchmark problems of 13 challenging high-dimensional MaOPs, the numerical results in terms of HV and IGD indicators demonstrate that MaOEA/D-CME achieves competitive advantages compared to some state-of-the-art MOEAs.